feat: adapt rapidocr v2

feat: adapt rapidocr v2

Author: SWHL

demo.py

@@ -10,41 +10,41 @@
 img_path = "tests/test_files/ch_en_num.jpg"
 
 # 方式一：使用base64编码传
-stime = time.time()
+stime = time.perf_counter()
 with open(img_path, "rb") as fa:
     img_str = base64.b64encode(fa.read())
 
 payload = {"image_data": img_str}
-response = requests.post(url, data=payload)  # , timeout=60
+response = requests.post(url, data=payload, timeout=180)
 
 print(response.json())
-etime = time.time() - stime
-print(f"用时:{etime:.3f}秒")
+elapse = time.perf_counter() - stime
+print(f"用时:{elapse:.3f}秒")
 
 print("-" * 40)
 
-# # 方式二：使用文件上传方式
-# stime = time.time()
-# with open(img_path, "rb") as f:
-#     file_dict = {"image_file": (img_path, f, "image/png")}
-#     response = requests.post(url, files=file_dict)  # , timeout=60
-#     print(response.json())
-
-# etime = time.time() - stime
-# print(f"用时:{etime:.3f}秒")
-# print("-" * 40)
-
-# # 方式三：控制是否使用检测、方向分类和识别这三部分的模型； 不使用检测模型:use_det=False
-# stime = time.time()
-# img_path = "../python/tests/test_files/test_without_det.jpg"
-
-# with open(img_path, "rb") as f:
-#     file_dict = {"image_file": (img_path, f, "image/png")}
-#     # 添加控制参数
-#     data = {"use_det": False, "use_cls": True, "use_rec": True}
-#     response = requests.post(url, files=file_dict, data=data)  # , timeout=60
-#     print(response.json())
-
-# etime = time.time() - stime
-# print(f"用时:{etime:.3f}秒")
-# print("-" * 40)
+# 方式二：使用文件上传方式
+stime = time.perf_counter()
+with open(img_path, "rb") as f:
+    file_dict = {"image_file": (img_path, f, "image/png")}
+    response = requests.post(url, files=file_dict, timeout=180)
+    print(response.json())
+
+elapse = time.perf_counter() - stime
+print(f"用时:{elapse:.3f}秒")
+print("-" * 40)
+
+# 方式三：控制是否使用检测、方向分类和识别这三部分的模型； 不使用检测模型:use_det=False
+stime = time.perf_counter()
+img_path = "tests/test_files/ch_en_num.jpg"
+
+with open(img_path, "rb") as f:
+    file_dict = {"image_file": (img_path, f, "image/png")}
+    # 添加控制参数
+    data = {"use_det": False, "use_cls": True, "use_rec": True}
+    response = requests.post(url, files=file_dict, data=data, timeout=180)
+    print(response.json())
+
+elapse = time.perf_counter() - stime
+print(f"用时:{elapse:.3f}秒")
+print("-" * 40)

rapidocr_api/main.py

@@ -3,7 +3,6 @@
 # @Contact: [email protected]
 import argparse
 import base64
-import importlib
 import io
 import os
 import sys
@@ -14,17 +13,7 @@
 import uvicorn
 from fastapi import FastAPI, Form, UploadFile
 from PIL import Image
-
-if importlib.util.find_spec("rapidocr_onnxruntime"):
-    from rapidocr_onnxruntime import RapidOCR
-elif importlib.util.find_spec("rapidocr_paddle"):
-    from rapidocr_paddle import RapidOCR
-elif importlib.util.find_spec("rapidocr_openvino"):
-    from rapidocr_openvino import RapidOCR
-else:
-    raise ImportError(
-        "Please install one of [rapidocr_onnxruntime,rapidocr-paddle,rapidocr-openvino]"
-    )
+from rapidocr import RapidOCR
 
 sys.path.append(str(Path(__file__).resolve().parent.parent))
 
@@ -39,36 +28,29 @@ def __init__(self) -> None:
             self.ocr = RapidOCR()
         else:
             self.ocr = RapidOCR(
-                det_model_path=det_model_path,
-                cls_model_path=cls_model_path,
-                rec_model_path=rec_model_path,
+                params={
+                    "Det.model_path": det_model_path,
+                    "Cls.model_path": cls_model_path,
+                    "Rec.model_path": rec_model_path,
+                }
             )
 
     def __call__(
-        self, img: Image.Image, use_det=None, use_cls=None, use_rec=None, **kwargs
+        self, ori_img: Image.Image, use_det=None, use_cls=None, use_rec=None, **kwargs
     ) -> Dict:
-        img = np.array(img)
-        ocr_res, _ = self.ocr(
+        img = np.array(ori_img)
+        ocr_res = self.ocr(
             img, use_det=use_det, use_cls=use_cls, use_rec=use_rec, **kwargs
         )
 
-        if not ocr_res:
+        if ocr_res.boxes is None or ocr_res.txts is None or ocr_res.scores is None:
             return {}
 
         out_dict = {}
-        for i, dats in enumerate(ocr_res):
-            values = {}
-            for dat in dats:
-                if isinstance(dat, str):
-                    values["rec_txt"] = dat
-
-                if isinstance(dat, np.float32):
-                    values["score"] = f"{dat:.4f}"
-
-                if isinstance(dat, list):
-                    values["dt_boxes"] = dat
-            out_dict[str(i)] = values
-
+        for i, (boxes, txt, score) in enumerate(
+            zip(ocr_res.boxes, ocr_res.txts, ocr_res.scores)
+        ):
+            out_dict[i] = {"rec_txt": txt, "dt_boxes": boxes.tolist(), "score": score}
         return out_dict
 
 
@@ -100,6 +82,7 @@ def ocr(
             "When sending a post request, data or files must have a value."
         )
     ocr_res = processor(img, use_det=use_det, use_cls=use_cls, use_rec=use_rec)
+
     return ocr_res
 
 

setup.py

@@ -77,9 +77,4 @@ def get_readme():
             f"{MODULE_NAME}={MODULE_NAME}.main:main",
         ],
     },
-    extras_require={
-        "onnx": ["rapidocr-onnxruntime"],
-        "paddle": ["rapidocr-paddle"],
-        "openvino": ["rapidocr-openvino"],
-    },
 )